Leaf Disease Identification: Enhanced Cotton Leaf Disease Identification Using Deep CNN Models

Agriculture plays a vital role in providing main source of food, income and employment to the rural people in economically developing countries. The major influencing factor which affects agriculture productivity is crop loss due to the plant diseases, which affects the production approximately 20 to 30%. To avoid such losses, conventional method has been done to identify the diseases but it is not accurate. Early and exact diagnosis of plant diseases is very important to avoid such losses caused by such diseases. But due to lack of proper cultivating knowledge, experience, and sense of disease prediction, sometimes those harvests and grains get harmed mostly or even totally. Obviously, that winds up with an enormous misfortune for the farmers and also for the financial development of the country. Thus, this paper tends to combine a piece of agriculture area with the help of Artificial Intelligence to reduce the loss due to infections of plant leaves. In order to solve this problem, we used the transfer learning models constructed with various CNN architectures like ResNet50, VGG19, InceptionV3, and ResNet152V2. We did experiments with these four methods on the standard cotton leaves dataset, to know which method gives the better performance in identifying cotton leaf diseases. Experimental results show that ResNet50, VGG19, InceptionV3, and ResNet152V2 are giving 75.76%, 87.64%, 96.46%, 98.36% respectively. Among the four models ResNet152V2 with parameters 60,380,648 gave more accuracy. So, this idea of using transfer learning method called ResNet152V2 for disease detection in plant is very useful and also gives more accuracy.